For David Millier

Started 3 months ago | Discussion thread
 Like?
 Re: I'll get my coat ;-) In reply to xpatUSA, 3 months ago

xpatUSA wrote:

Since everybody is still talking about frequency in spite of my incredibly lucid post earlier, I'll just stand on the sidelines and leave y'all with some light reading:

http://www.dspguide.com/ch25.htm

Its a quite mixed bunch at the battlefield today

Almost all have something interesting to say. Something to learn from.

But, it also seem that most have just set their communication mode to output.

There are two very different and constructive ways of evaluating the sensor design.

• One is strictly scientific, where you base your arguing on sound theories and/or thorough measurements.
• One is more subjective. What can you or what can other good photographers do with the tool. Can you show that this tool gives you means for making superior results.

But, I think both misses the mark in discussions like this.

Take the sampling theorem discussions. They never get anywhere. Partly because the same "facts" are repeated over and over. But the "facts" mentioned are simplified and also sometimes wrong.

The sampling theorem is based on three fundamentals

1. You need a steep low pass filter at half the sampling frequency
2. You need to do a noise free point sampling
3. You need an elaborated reconstruction method

None of those three are implemented in any camera, except maybe in astronomical telescopes or other scientific instruments. In particular, number three is totally ignored. Number three is based on fourier analysis. You need to use some fourier series to do the reconstruction. In practice this means that if you e.g. sample at 10 um spacing, you have to reconstruct the sinus curves at maybe a resolution of 1 um. Otherwise you cannot get a distortion free sine wave for the highest sampling frequency. This means that if you have a 15 MP sensor, then the resulting image should be 100 times larger, i.e. 1.5 GP.

OK - back to reality. We dont have 1.5 GP sensors. At least not in our consumer cameras.

But - as a message to those that claim we need to consider the sampling theorem - dont you forget that the sampling theorem needs all three above.

--
/Roland
X3F tools:
http://www.proxel.se/x3f.html
https://github.com/rolkar/x3f

Complain
Post ()
Keyboard shortcuts: